Rust Statsd Server

This is a statsd server implementation written in rust. It aims to be as compatible as possible with etsy/statsd.

Building the application

1. Clone the repository.
2. Run make install.
3. The generated binary will be in dist/statsd.

Running tests

After cloning the repository you can run either the unit tests, and or the integration tests. The integration tests require python 2.7 and that port 8125 be free on your system.

make test

Will run both the unit and integration tests.

Usage

The statsd server has several options to control which ports it runs on:

-p, --port=<p>        The UDP port to bind to [default: 8125].
--admin-host=<p>      The host to bind the management server on. [default: 127.0.0.1]
--admin-port=<p>      The port to bind the management server to. [default: 8126]

Changing how frequently metrics are output

--flush-interval=<p>  How frequently to flush metrics to the backends in seconds. [default: 10].

On each flush interval event, derived metrics for timers are calculated. This duration is tracked as statsd.processing_time. You can use this metric to track how long statsd is spending generating derived metrics.

Enabling the console or graphite backends

By default no backends are enabled. In this mode the statsd server doesn't do that much. To enable one or both backends use the CLI flags:

--console             Enable the console backend.
--graphite            Enable the graphite backend.

The graphite backend has additional options for defining where graphite/carbon runs:

--graphite-port=<p>   The port graphite/carbon is running on. [default: 2003].
--graphite-host=<p>   The host graphite/carbon is running on. [default: 127.0.0.1]

Internal metrics

This server tracks a few internal metrics:

• statsd.bad_messages The number of invalid metrics that have been sent since the last flush.
• statsd.total_messages The number of messages received including invalid messages.
• statsd.processing_time How many ms were spent calculating derived metrics in the current flush cycle.

Prior Art

I took a bunch of inspiration in how to implement and structure this implementation from erik/rust-statsd.